phi3-offline-dpo-lora-noise-0.0-5e-5-42

This model is a fine-tuned version of microsoft/Phi-3-mini-4k-instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6559
• Rewards/chosen: -0.1041
• Rewards/rejected: -0.1884
• Rewards/accuracies: 0.7341
• Rewards/margins: 0.0842
• Logps/rejected: -402.5698
• Logps/chosen: -418.9837
• Logits/rejected: 12.3730
• Logits/chosen: 14.0651

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.691	0.1778	100	0.6815	-0.0320	-0.0641	0.5913	0.0320	-390.1401	-411.7723	11.8762	13.7530
0.6908	0.3556	200	0.6608	-0.1168	-0.1872	0.7262	0.0704	-402.4569	-420.2538	12.8524	14.4928
0.6936	0.5333	300	0.6605	-0.0593	-0.1315	0.7222	0.0722	-396.8808	-414.4962	12.5717	14.2749
0.6935	0.7111	400	0.6625	-0.0799	-0.1500	0.7460	0.0701	-398.7299	-416.5561	12.4923	14.1862
0.6941	0.8889	500	0.6554	-0.1087	-0.1935	0.7421	0.0847	-403.0800	-419.4448	12.3764	14.0679

Framework versions

• PEFT 0.7.1
• Transformers 4.42.3
• Pytorch 2.3.0+cu121
• Datasets 2.14.6
• Tokenizers 0.19.1